Research
A paper is listed when one of its authors is affiliated with the lab. Data: OpenAlex.
SubRot: Signed Gradient Subspace Calibration for VLM Rotation QuantizationiFlytek · arXiv (Cornell University)Sep 28Rubric Rewards from Item Response TheoryMicrosoft · arXiv (Cornell University)Sep 28PEARL: Adaptive Prefill-Decode Execution with Elasticity for Agentic Reinforcement LearningAlibaba · arXiv (Cornell University)Sep 28Beyond One Epoch: Uncertainty-Weighted Sensitivity Regularization for Recommendation ModelsMeta · arXiv (Cornell University)Sep 28Reward-Aligned Reweighting for On-Policy DistillationAlibaba · arXiv (Cornell University)Sep 28Shockingly Simple Self-retrospection Improves Agentic Models Without RLMicrosoft · arXiv (Cornell University)Sep 28Rethinking Visual Token Compression for Video Large Language Models: A Simple Yet Strong BaselineSenseTime · arXiv (Cornell University)Sep 28Token-Disentangled Latent Test-Time Scaling for Vision-Language ReasoningAlibaba · arXiv (Cornell University)Sep 28Muon Sublates the Edge of Stability in LLM PretrainingAlibaba · arXiv (Cornell University)Sep 28Planarian: Managing Agent State with StatepointsMicrosoft · arXiv (Cornell University)Sep 28Reinforcing Agentic Creativity in Scientific Ideation with Night ScienceMicrosoft · arXiv (Cornell University)Sep 28Don't Let AI Agents YOLO Your Files: Information and Control in Agent-Native FilesystemsMicrosoft · ACM SIGOPS Symposium on Operating Systems Principles (SOSP)Sep 28Indistinguishability of Sum of Permutations: A Fourier Analytic Route to Classical and Quantum SecurityTechnology Innovation Institute · arXiv (Cornell University)Sep 28Rotated Manifold Optimization for Low-Rank AdaptationMicrosoft · arXiv (Cornell University)Sep 28TSGate: Timestep-Aware Gated Attention for Diffusion TransformersAlibaba · arXiv (Cornell University)Sep 28When Confidence Rises Too Early: Detecting Shortcut Reasoning via Premature Answer CommitmentAlibaba · arXiv (Cornell University)Sep 28QProf: Fleetwide Transitive Cost Profiling of Warehouse-Scale ServicesAlphabet · ACM SIGOPS Symposium on Operating Systems Principles (SOSP)Sep 28Validating a High-Performance Cloud Object Store with Lightweight Formal MethodsMeta · ACM SIGOPS Symposium on Operating Systems Principles (SOSP)Sep 28Correcting to Predict: Pseudo-Value Correction for Multimodal Attribute Value ExtractionAlibaba · arXiv (Cornell University)Sep 28Efficient World Action Model Inference with Adaptive Intermediate StatesAlibaba · arXiv (Cornell University)Sep 28AUV-Bench: Aesthetic Understanding and Generation Evaluation for User InterfacesAlibaba · arXiv (Cornell University)Sep 28LLMs as Adaptive Meta-Solvers: Strategy-Diverse RL for Industrial-Scale OptimizationAlibaba · arXiv (Cornell University)Sep 28Dynamic Flow, Static Graph: KV Cache Reuse for Efficient LLM Serving on Mobile NPUsAlibaba · arXiv (Cornell University)Sep 28SIPO: Selective-Inference Policy Optimization for Tree-Structured Agentic RLMininglamp Technology · arXiv (Cornell University)Sep 28Long-Horizon Scaling: How Model Capabilities Shape the Returns to ComputationAllen Institute for AI · arXiv (Cornell University)Sep 28